Ajith Kumar
Abstract—Large Screens are widely used in present day applications, especially in areas like information visualization, real-time
monitoring of large systems, Geospatial information systems etc. Large Screens improve the user performance on these systems.
By performance we refer to the speed in which the users can carry out their tasks. There are several factors that affect the user
performance. Firstly the type of hardware device used in the setup, secondly the possibility of the configuration used in creatively
exploiting the human abilities. Human abilities include physical navigation and peripheral vision. Ultimately the performance of the
system used is still dependent on the corresponding application for which it is used. It is a challenge to come up with an optimum
solution for each application. This survey is to study some of these factors using given set of taks and their results. Using the
inferences obtained from the results to give a set of rules while using these types of display devices. We also discuss on the topic of
the effect of bezels on visualization tasks.
1 OVERVIEW
Visualizing large sets of data on a single user desktop can be a tedious
task. A simple overview of such large data set might not fit into the
screen of such a desktop. In such case, user would have to pan and
zoom several times to study the data carefully. In spite of immense
effort used, it is still difficult to get a clear overview of the data. This
is expensive in terms of time consumed and also on frustration caused.
One of the solutions is to use large displays. There has been a much
work in this area and different possible hardware configurations, for
building large displays have been proposed. One of the possibilities is
to use a single large screen, which was then was an expensive solution.
Another solution is, to use the smaller displays and to combine them to
obtain one large screen. This resulted in construction of displays like
the Tiled display [1] . The tiled display can also be built in other ways.
A tiled displays is further classified into two types flat and curved dis-
plays. These experiments yield some interesting results. Some other
types of flat displays are also compared namely the focus+context dis- Fig. 2. Approach
play and overview+detail display.
2 I NTRODUCTION
In this section we briefly introduce the various types of displays used
and their respective features.
3 A PPROACH
In this section we briefly describe the approach of the survey. First step
Fig. 5. Focus+context display[4] is to study the need for large displays. We then move-on to the types
of large displays present and their features. Among the types of large
displays present, the ones which have the potential usefulness both
2.3 Human vision and Physical navigation in terms of cost and effectiveness are chosen. The tasks selected to
compare the displays and the experimental setup used for the same are
2.3.1 Peripheral Vision and focus studied carefully. The analysis and results of these experiments gives
Peripheral vision or the side vision is the ability to see objects and some insight on the salient features and differences of the displays
movement outside the direct line of vision. Considering the high reso- compared. Using the inference from these set of analysis and results
overview+detail display. The overview+detail display had one mon-
itor dedicated for detailed data and the other one for overview of the
entire data.
4.3 Results
The results obtained from first three experiments was that the curved
display seemed to outperform the flat display in case of the search
task and the comparison task. This is due to the fact that in curved
display setup translational movement required is less compared to the
flat display setup. For the insight task Flat displays where preferred
over the curved display. The reason was that the in flat display users
had the provision to step back and get an overview of the display and
see things in context. The curved display is built such that when user
steps back and looks at the display, users get a distorted image because
Fig. 8. Approach
of the angle in which the displays are placed.
The last three experiments where implemented on the fo-
cus+context display and the overview+detail display. In the paper
we give out the important details to be considering while choosing the referred zoom+pan display which uses a single desktop is also con-
large displays. The pipeline chosen is as shown in the figure 8 sidered, but in the result point of view the its stance is quite negligible
compared to the other two displays. The results of the first two experi-
4 E XPERIMENTS S TUDIED ments shows, that the time consumed by the focus+context is less than
4.1 Tasks overview+detail display with a decent margin. In the car task the fo-
We have picked up six basic visualization tasks among the several task cus+context display completely outperforms the overview+detail dis-
studied during the survey. Some of the tasks are based on Geo Spatial play. This says that in dynamic views where required concentration
Information. The first three of the tasks where used to compare flat level is higher focus+context display setups work better. Both these
displays and curved displays. The last three tasks are for the compar- types of displays and these experiments entirely rely on the peripheral
ison between overview+detail display and focus+context display. The vision concepts.
first five tasks are static tasks. The last one is a dynamic task, used to
get insight on dynamic views. The tasks are: 5 B EZELS AND ITS EFFECTS
Bezels as described earlier refers to physical discontinuities on large
• Search task : The user has to search a particular icon or figure screens. Bezels can be seen in tiled displays. There were quite good
over the given map. number of study on these topic. Bezels [5] always have an effect on
the user performance. The particular scenarios where Bezel effects
• Comparison task : The user was asked to compare the data of can be seen are:
given two locations over the map.
• Search task: Search task is affected by the presence of bezels
• Insight : The basic idea of visualization is to obtain insights on when the target to be searched is split across the Bezels [9],
the given information. So users where asked to give out the first which makes it difficult to identify the target. It can also be ad-
insight they obtained from the map. vantageous in terms of dividing the search space into grids, there-
fore making it easier to the user to remember the space searched
• Circuit board task: Users are presented with an image of a circuit
previously. In the alternative case where single large screens are
board of some device. The task was to verify the circuit connec-
used users might have to virtually separate the search space but
tions by matching any source connection on to the chip placed in
virtually separating spaces could be quite tedious sometimes.
the centre of the screen. There were correct corrections as well
as incorrect ones. • Map navigating task: Map navigating refers to finding a path be-
tween a source and a destination on the map. In this case bezels
• Map task: Similar to the search task, here the users have to find
might affect negatively due to the discontinuity of the maps be-
the nearest hotel to a given location, nearest in terms of time to
tween the bezels and distorted effect on the map.
travel.
6 I NFERENCES FROM THE STUDY
• Car task: In this task the user has to drive a car through a road
amidst the mountains. The constraints here are that nails are 6.1 Bezels
placed on the road, which are clearly visible in the focus/detail The presence of bezels has both merits and demerits. The actual effect
region and the user has to prevent the car from running over them is more application dependent. Basically the static tasks such as search
by changing the direction. The second constraint was to avoid tasks are still less effected by the bezels. The bezels can be more
stones falling from the mountains. The stones falling could be distracting when dynamic scenarios are considered. Especially in high
seen only in the context/overview part of the display. resolution gaming scenarios [7], users feel distracted from the bezels
due to distorted view as shown in the figure 9. More the bezel size
4.2 Experimental Setup more distraction is felt. Hence in these cases rear projectors setup is
The first 3 tasks defined where run on a setup containing a display of usually preferred by users. In Few cases the problem of bezels can be
size 24 panels and 31.5 mega pixel resolution. The display was used solved to some extent by creating extra software to manage the display.
in both flat as well as curved setup. In the flat setup the 24 panels
where placed adjacent to each other in a grid structure using 3 panels Mackinlay et.al [6] suggested a way in which any unnecessary data
vertically and 8 rows horizontally. Each panel was a 17 inch LCD present in the vicinity of the bezels, can be drawn at the place of bezels
with the highest resolution of 1280 × 1024 . Also a 12 GNU/Linux hence omitted in the actual screen retaining continuity in the images.
computers to power the display.The same displays where used for the This approach can be used in some cases where some low priority
curved setup but the panels where placed at an angle of 30deg with the data for example in graph of links and nodes the links near the bezel
adjacent panel. are can be compromised. The graph with nodes split across the bezels
The last 3 tasks used large screens(any of the two types) is shown in the figure 10. The graph obtained after compromising the
for focus+context displays and tiled displays with 2 monitors for links in the vicinity of bezels is shown in the figure 11. This does
tasks. This is mainly due to the reduced need for translational move-
ment and more number of pixels directly visible to the user at a given
time. Hence for tasks which involve looking throughout the display
or searching something on it Curved displays can be preferably used
over the flat display. But when it comes to the insight task the curved
display is less preferred by the users. This is due to the fact that there
is no provision to step back in the curved display. Stepping back will
give only a distorted image in this case. So whenever an application
requires to be looked in to get visualization insights Flat displays come
into picture.
6.2.2 Flat display
Even though flat display setups are outperformed by the curved display
Fig. 9. This shows the distorted effect caused by the bezels [6]
it is still required in some important steps of visualization pipelines.
Precisely in the task to gain important insights in the visualized data.
The flat displays provide room for all possible physical navigation
not work in the same manner for geo-spatial data with maps involved. such as translational movement and rotational movement. This pro-
Applying it on maps would cause some roads, street names or some vides the users an option to move back and forth to get overview look
other data to be lost, which is not acceptable. We conclude on this say- into more data together and get a proper image to look at think about
ing that any large volume data with some low priority data with it can the possible insights.
be approximated using the method given by Mackinlay et.al to obtain
better imagery effects. The effect of bezels on the graph example is 6.3 Combining the curved Display and the flat Display
given in the following figures. Given the detailed advantages of both curved display and flat display
it is clearly understood that both of these are required in some stage of
the visualization pipeline. This is because visualization pipeline will
always have a task to gain insights to the given data and it will be of
prime importance. Hence it is good to achieve a setup which combines
curved and flat displays. As the possible solution mentioned by Shupp
et.al [8] it would be good to have a setup that uses tiled monitors and
can be used as both curved display as well as flat display as and when
required. Thus obtaining the benefits of both types of displays.